Per-dataset row signatures of tables already in a connection

The local twin of publish_object_signatures(), for a publisher that materializes the core before building (the OBIS and netCDF publishers do, to join across tables): each table's rows grouped by dataset_key and signed. A table with no dataset_key column is signed per dataset through via — the rows a dataset actually reaches by a key — so a vocabulary row another dataset added does not rebuild this one; without a via it is signed whole (dataset_key = "*"). A table absent from the connection contributes nothing, which publish_data_parts() reports as "<absent>".

Usage

publish_table_signatures(con, tables, via = list())

Arguments

con

a DuckDB connection holding the tables

tables

table (or view) names

via

named list, table -> c(from, key): sign table's rows joined to the distinct (dataset_key, key) pairs of table from, e.g. list(taxon = c("obs_bio", "taxon_key"), cruise = c("sample", "cruise_key"))

Value

The data frame publish_object_signatures() returns, content_hash empty.